karaf-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Guillaume Nodet <gno...@apache.org>
Subject Re: Transitive feature/bundle dependencies
Date Mon, 24 Feb 2014 10:46:05 GMT
I agree, I think the feature is broken, as the bundle should not be flagged
as dependency = true if it's supposed to be installed.

Currently, the resolvers we have always resolve bundles feature by feature,
i.e. the resolver is only used to compute the bundles needed for a given
feature, and not as a whole.
I think that's ok, though at some point, I'd like to think about
introducing an enhanced resolver (using the real osgi resolver) that could
actually resolve features in a single pass, in which case, the dependencies
bundle would be transitive dependencies.

Guillaume


2014-02-24 9:45 GMT+01:00 Grzegorz Grzybek <gr.grzybek@gmail.com>:

> Hello
>
> I come to conclusion, that rather the feature definition below is
> incorrect, not the feature:install command:
>
> If someone wants to install this feature *alone*, he/she would end up with
> installed feature, but *not* with installed bundle:
>
>     <feature name="jclouds-api-rackspace-
> cloudidentity" description="Rackspace Cloud Identity API"
> version="1.6.2-incubating" resolver="(obr)">
>         <feature version='1.6.2-incubating'>jclouds-compute</feature>
>         <feature
> version='1.6.2-incubating'>jclouds-api-openstack-keystone</feature>
>         <bundle
>
> dependency='true'>mvn:org.apache.jclouds.api/rackspace-cloudidentity/1.6.2-incubating</bundle>
>     </feature>
>
> so I think it's better to change the feature definition, not the loader
> process.
> By the way: it's related to https://issues.jboss.org/browse/ENTESB-1209
>
> regards
> Grzegorz Grzybek
>
>
> 2014-02-18 12:55 GMT+01:00 Ioannis Canellos <iocanel@gmail.com>:
>
> > I think that the original intention was not to have the dependency
> > attribute on jclouds-api-rackspace-cloudidentity. Most probably this
> > was a side effect of splitting a feature into multiple features.
> >
> > Of course, the problem that you describe may apply to other cases too.
> > So, I think that the change you propose makes sense.
> >
> > Other thoughts?
> >
> > --
> > Ioannis Canellos
> >
> > Blog: http://iocanel.blogspot.com
> > Twitter: iocanel
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message